The rapid adoption of robotic-assisted total knee arthroplasty (RATKA) has resulted in pre-operative CT scans becoming more readily available. After the segmentation and identification of landmarks by trained segmentation specialists, the Mako SmartRobotics^TM software generates measurements of interest for the calculation of the arithmetic hip-knee-ankle axis (aHKA), joint line obliquity (JLO), and the Coronal Plane Alignment of the Knee (CPAK) phenotype. The aim of this study is to ascertain how closely correlated these two sets of readings are and whether the CPAK distribution is altered when comparing both modalities. A retrospective radiological study was undertaken on 500 knees (367 patients: 133 bilateral, 234 unilateral) comparing the CT-based software-generated measurements of patients undergoing RATKA using the Stryker Mako system against manual measurements derived from long limb radiographs (LLRs). There were statistically significant differences between the average measurements of the LDFA (0.27 ± 2.95, p = 0.045), MPTA (1.15 ± 2.20, p < 0.001), aHKA (1.41 ± 3.85, p < 0.001) and JLO (0.89 ± 3.50, p < 0.001), with CT measurements having higher mean readings for LDFA, lower readings for MPTA, more varus aHKA and increased apex distal JLO. Despite this, correlation was moderately good: LDFA (r = 0.409, p < 0.001), MPTA (r = 0.683, p < 0.001), aHKA (r = 0.595, p < 0.001) and JLO (r = 0.456, p < 0.001). The CPAK distribution was also significantly different. LLRs underestimate the degree of constitutional varus and JLO compared to CT-based software-generated measurements, with a resultant increase in CPAK Types I and IV when using CT measurements. Despite moderately good correlation between both imaging modalities, there remains a statistically significant difference between them.